The present disclosure relates to a technique for implementing data loading in hierarchical pipeline for maximum utilization of computer resources (like GPU, etc.) and to obtain effective prediction performance from the numerical and categorical features. The technique involves dividing the data into chunks of data, batching them to form tensors and loading the batched data to the framework to be executed on the GPU. The framework generates embeddings for the numerical and categorical features identified from the batched data and determines the interactions between the generated embeddings to provide a predictive output.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.